Solved: Asus Rog Strix 3090 (non-oc) TimeSpy GPU score... lowish?

KrazeyKami

n00b
Joined
Dec 31, 2010
Messages
61
Solved: nothing wrong with the card. The Asus Rog Strix 3090 (non-oc) has a stock boostclock of 1695 mhz @ 350 TDP. This is low compared to other cards (e.g MSI Trio) which have a stock boost of 1785 mhz @ 370 TDP. It explains the difference in scores and reviews.
After adding 165 mhz Core clock and 112% Powerlimit (on par with Strix OC version), much better results: https://www.3dmark.com/3dm/70540804 (GPU score 20367).


Hello everyone :)

I'm the proud owner of an Asus Rog Strix 3090 (non-OC version).
However, after some benchmarking, I feel my 3D Mark results are sub-average:

Default Mode: https://www.3dmark.com/spy/25366699 (GPU score 18 811)
OC Mode: https://www.3dmark.com/spy/25367015 (GPU score 19 277)

Shouldn't my results be closer to 20k GPU score?

A friends' 3080 Ti (FE) scores 19 307 at stock / default settings (https://www.3dmark.com/3dm/70096477).
Granted it's not the same card, but still...

Also, I find theres a rather large gap between my clock and average clock frequency, compared to similar cards:

Clock frequency: 1.965 MHz
Average clock : 1.736 MHz

Other cards /w same specs

While I can squeeze more performance out of the card when overclocking, it still yields much lower results.
I'm actually looking for an explanation why it is below average in stock mode (I'm in the low 10-20%).
The best I can squeeze out of it is 20609 GPU score, and that is with a high OC (https://www.3dmark.com/spy/25296286).

Especially compared to the premium price for this card, I wouldn't expect to see 3080s get better results at a much lower price.

Are there any other owners of this exact card (Asus Rog Strix 3090) who can share their stock results?

Many thanks for your insights!

P.S.,
I tried the following:
- DDU / reinstall drivers
- Reinstall Windows and only install graphics driver / no Asus bloatware / no overclock software
- Run bench in Safe mode
- Set Bios PCI to Gen 4
- Enable / Disable BAR
- Other things i cannot remember ^^

gpu-z.JPG

cpu.jpg

mem.jpg

mobo.jpg

spd.jpg
 
Last edited:
Could be temperature related seeing the clocks are lower then expected.
 
I did think of that as well, but my GPU temp is around 55c, and memory between 60-70c when benchmarking. Rather low, but then again, I've got some good cooling in my case.
Also GPU-Z never mentions thermal throttling as a perf cap, always power limit...

Tomorrow I'm receiving my AIO waterblock for the card. I'll try again and post the scores / temps. Good to see atleast y'all think that the avg. clock shouldn't be dropping like that.

Thanks so far!
 
You using 3 different PCIe power cables? Using 2 connectors on 1 cable can cause some issues with high power draw, or maybe the PSU you have is not up to it.
 
Hi Denpepe,

Yep, using 3 separate cables. I have a 1500 watt PSU, so I'd think it's fine.

I've installed the AIO Alphacool Eiswolf 2 waterblock today.
It lowered my GPU temperature to 25c idle, 48c when benchmarking.
Vram is between 35 (idle) and 56 (bench), while memory junction is 38c (idle), peaks at 82c (bench).

I re-ran the benchmark, but the results aren't much better. In fact, not only does my GPU seem to run at a lower average still, but now also my memory seems to go lower average...

https://www.3dmark.com/compare/spy/25412783/spy/25366699

Not sure what to make of all this...

stats.jpg
 
Could always oc the card. Ton of variables from background processes to Temps to sicon lottery.
 
gpuz shows a boost clock of 1695, he says hes getting 1736 so i dont think its that. not sure where the 1965 came from. maybe its a cpu or ram speed difference? hes only at 4.5 and 2400. its close...
He should be boosting to somewhere in the 1900mhz range and if temp/voltage are fine it should consistently stay in range. And those look fine to me.

OP you turned off gsync/freesync (+vsync) for your benchmarks right?
 
He should be boosting to somewhere in the 1900mhz range and if temp/voltage are fine it should consistently stay in range. And those look fine to me.

OP you turned off gsync/freesync (+vsync) for your benchmarks right?
ah, i see now. its supposed to be 1860/1890 depending on mode. odd that gpuz would report it low, somethings up.


op, this a new card?
 
ah, i see now. its supposed to be 1860/1890 depending on mode. odd that gpuz would report it low, somethings up.


op, this a new card?
1695 is the rated boost clock for a non oc 3090. They boost higher in reality but that is the rating.
 
Thanks for your replies so far.
Furious_Styles: Yep, G-Sync is turned off.
Pendragon: It's a new card, bought it in March last year (2021).
mnewxcv: I've done a undervolt following your guide (thanks for that!).

My results (2040 curve) @ stock with the undervolt / 100% power: https://www.3dmark.com/spy/25413357. GPU score: 19310 (nearly 1000 points lower than yours).
Would it matter that I'm running a 3840x1600 screen (g-sync off)? I believe it shouldn't matter with Timespy, altho I did notice a little improvement when running it from my second screen (2560x1440).

When running @ 137% power, I get better scores, but still feels lowish considering the card & settings: https://www.3dmark.com/spy/25413717. GPU score: 20 804
It's stable as well in Bright, but I didn't want to push any higher as my memory junction temp was reaching 96c.
I did get stutter though on that Bright benchmark, I suppose that isn't normal? :unsure:
Edit: no stutter in Win11.
 
Last edited:
Thanks for your replies so far.
Furious_Styles: Yep, G-Sync is turned off.
Pendragon: It's a new card, bought it in March last year (2021).
mnewxcv: I've done a undervolt following your guide (thanks for that!).

My results (2040 curve) @ stock with the undervolt / 100% power: https://www.3dmark.com/spy/25413357. GPU score: 19310 (nearly 1000 points lower than yours).
Would it matter that I'm running a 3840x1600 screen (g-sync off)? I believe it shouldn't matter with Timespy, altho I did notice a little improvement when running it from my second screen (2560x1440).

When running @ 137% power, I get better scores, but still feels lowish considering the card & settings: https://www.3dmark.com/spy/25413717. GPU score: 20 804
It's stable as well in Bright, but I didn't want to push any higher as my memory junction temp was reaching 96c.
I did get stutter though on that Bright benchmark, I suppose that isn't normal? :unsure:
if you want to compare timespy results you have to use the stock settings. so if youve changed it to 3840x1600, lower it to the stock 1440p or whatever it is.
 
How old is your windows install, and do you have a lot of software installed?
 
1695 is the rated boost clock for a non oc 3090. They boost higher in reality but that is the rating.
ah, i see. he says hes at 2040 now so that seems normal BUT hes* testing at an odd res which might skew the numbers.
edit *might, but i guess not.
 
Last edited:
ah, i see. he says hes at 2040 now so that seems normal BUT hes testing at an odd res which might skew the numbers.
Hi Pendragon, no, I didn't alter the resolution in TimeSpy. It's the default. My monitor is a 3840 x 1600 though. Afaik, it shouldn't matter as it gets upscaled. The benchmark is default.

How old is your windows install, and do you have a lot of software installed?
Fairly recent, not many applications installed and definitly not much running (I close everything that's running in background etc).
 
His score looks perfectly fine. Plenty of other examples for reference now that I've had a sec to look at it.

https://hardforum.com/threads/rtx-3090-h-owners-official-3dmark-time-spy-leaderboard.2003111/
Interesting that post :) I read it a couple of days ago, and yes, it gave me some piece of mind.

So, the whole 'issue' here is that I've lost the silicon lottery, as where you had won it? ;)
Was thinking also about my RAM speeds... it shouldn't count to GPU score, but I read a lot of posts of people saying it does nonetheless. I'm at 3200 mhz (Hyper-X Octo).
It's probably nothing.

Overall, not saying my scores are that bad, but I felt it was a tad lower than average. My friend's 3080 Ti scores higher, which might be expected (or not). The reviews are not really consistent as to whether it should be a tad faster or slower than the normal 3090.

P.S.,
Interestingly, if I only use 1 monitor (the 3840x1600), it actually scores about 250 points higher.
 
Interesting that post :) I read it a couple of days ago, and yes, it gave me some piece of mind.

So, the whole 'issue' here is that I've lost the silicon lottery, as where you had won it? ;)
Was thinking also about my RAM speeds... it shouldn't count to GPU score, but I read a lot of posts of people saying it does nonetheless. I'm at 3200 mhz (Hyper-X Octo).
It's probably nothing.

Overall, not saying my scores are that bad, but I felt it was a tad lower than average. My friend's 3080 Ti scores higher, which might be expected (or not). The reviews are not really consistent as to whether it should be a tad faster or slower than the normal 3090.

P.S.,
Interestingly, if I only use 1 monitor (the 3840x1600), it actually scores about 250 points higher.
try setting your cpu bus speed and multiplier to normal too...
1641519594463.png
 
Thanks for your replies so far.
Furious_Styles: Yep, G-Sync is turned off.
Pendragon: It's a new card, bought it in March last year (2021).
mnewxcv: I've done a undervolt following your guide (thanks for that!).

My results (2040 curve) @ stock with the undervolt / 100% power: https://www.3dmark.com/spy/25413357. GPU score: 19310 (nearly 1000 points lower than yours).
Would it matter that I'm running a 3840x1600 screen (g-sync off)? I believe it shouldn't matter with Timespy, altho I did notice a little improvement when running it from my second screen (2560x1440).

When running @ 137% power, I get better scores, but still feels lowish considering the card & settings: https://www.3dmark.com/spy/25413717. GPU score: 20 804
It's stable as well in Bright, but I didn't want to push any higher as my memory junction temp was reaching 96c.
I did get stutter though on that Bright benchmark, I suppose that isn't normal? :unsure:
96C mem temp is nothing to worry about. If it goes above 105C then you need to change the fan profile perhaps.
 
Can you elaborate? Currently I have PBO and all settings with regards to the Bios on 'Auto'. The only thing I changed there is the memory profile (OCMP) so it runs 3200.
for some reason they are off. the multi being .25 is odd but works i guess and the bus should be 100.
 
Maybe you should try running Port Royale. It's more GPU bound and less system dependent than Time Spy.

Run it with the card at default. I would expect the score to be around 13k.
 
Maybe you should try running Port Royale. It's more GPU bound and less system dependent than Time Spy.

Run it with the card at default. I would expect the score to be around 13k.
https://www.3dmark.com/3dm/70483388

Well, atleast it's consistent with scoring below average :LOL:
Score is 12.8k, so on the ball with the 13k. Interesting again how the GPU decides to run at a 1785 clock speed. Atleast the average isn't far off this time... :whistle:
 
TimeSpy Extreme @ Default / Stock:
https://www.3dmark.com/spy/25428861

This is more like it... and rather how I'd expect the 'normal' TS to run as well. At any rate, I think I'm good. It's all marginal with a few deviations, and as mnewxcv mentioned, it's rather inline with other 'stock' results.

Thanks for the enlighting discussions, hopefully this helps someone else that's stressing around the scores as well (y)
 
https://www.3dmark.com/3dm/70483388

Well, atleast it's consistent with scoring below average :LOL:
Score is 12.8k, so on the ball with the 13k. Interesting again how the GPU decides to run at a 1785 clock speed. Atleast the average isn't far off this time... :whistle:

That's about right for default settings. Gotta love the marketing they do with these cards. I suspect the only real difference between your card and the OC version is the vBios. Your card has a default PL of 350w, but max of 480w just like the OC version. So in the end probably no real difference if you max out the PL.

Strix,
https://www.techpowerup.com/vgabios/228823/asus-rtx3090-24576-201117-1

Strix OC,
https://www.techpowerup.com/vgabios/224964/asus-rtx3090-24575-200909

So if you max out the PL and apply a fairly conservative OC (i.e. 120-150 core and 500-1000 mem) you'll probably hit or break 14k in PR just like any 3090. Most likely your relatively slow RAM kit is hurting the TS score. But that tends too only happen on benches/games that run over 100fps. So depends on what you play as to whether it matters at all.
 
To close the loop, I found the reason why reviews show 3090s that have slightly better scores than the 3080 Ti FE...:

My card (Strix 3090) has a stock boostclock of 1695 mhz @350 tdp. The cards used in those reviews are the MSI Trio version with a boostclock of 1785 mhz @370 tdp.. it adds up completely. And like you say talon95, when I apply a higher PL (up to 480 tdp for this model) and some moderate OC, the card is absolutely killing it. And with the Eiswolf 2 wb, the temps are very low, so there's no reason not to apply it.

Atleast I'm happy that theres nothing wrong with my card, so I can sleep again :p Hope this thread is helpful to others with similar concerns. Thanks again all for your help and thoughts!
 
Last edited:
That's about right for default settings. Gotta love the marketing they do with these cards. I suspect the only real difference between your card and the OC version is the vBios. Your card has a default PL of 350w, but max of 480w just like the OC version. So in the end probably no real difference if you max out the PL.

Strix,
https://www.techpowerup.com/vgabios/228823/asus-rtx3090-24576-201117-1

Strix OC,
https://www.techpowerup.com/vgabios/224964/asus-rtx3090-24575-200909

So if you max out the PL and apply a fairly conservative OC (i.e. 120-150 core and 500-1000 mem) you'll probably hit or break 14k in PR just like any 3090. Most likely your relatively slow RAM kit is hurting the TS score. But that tends too only happen on benches/games that run over 100fps. So depends on what you play as to whether it matters at all.
I've aligned with the OC specs; Simply added 165 mhz Core clock, and set the PL to 112% (390 TDP), and the results speak for itself:
TS: https://www.3dmark.com/3dm/70540804 (GPU score 20367)
PR: https://www.3dmark.com/3dm/70541325 (GPU score 13801)

🎉🥳
 
Last edited:
Back
Top